The influence of a schizophrenia polygenic risk score on white matter microstructure in schizophrenia, bipolar disorder and health

Abstract

Schizophrenia (SZ) and bipolar disorder (BD) are classified as psychiatric disorders by the Diagnostic and Statistical Manual of Mental Health Disorders fifth edition (DSM-5). Both illnesses are highly heritable, share symptomatology, and have a polygenic architecture. Given the polygenic architecture of these disorders, several genetic risk variants can be combined in a compound polygenic risk score (PRS) previously associated with the illness, and its impact can be tested on brain phenotypes that have also been associated with the illness, in a so called neuroimaging genetics approach. To further validate and better characterize a recently reported PRS for SZ, I have assessed its impact on white matter (WM) microstructure, which is known to be impaired in SZ, and to a lesser extent, in SZ relatives and in BD. Each participant of this study was genotyped in order for its SZ PRS to be calculated. The PRS used was compounded using the genetic risk variants found in the latest Psychiatric Genomics Consortium (PGC) schizophrenia meta-analysis study, and was found to explain almost 10% of the variance in SZ status. In the present study, I estimated the effect of this SZ PRS on brain WM microstructure, using two proxy features [fractional anisotropy (FA) and mean diffusivity (MD) ], previously collected with diffusion tensor imaging (DTI) in a Caucasian group of SZ (n=22), BD (n=25), relatives (REL; n=28) and healthy controls (HC; n=68). The DTI images of each participant were pre-processed and analysed with FSL software. Resultant of the pre-processing steps, FA and MD images were created. Using tract-based spatial statistics (TBSS) approach, the FA and MD images were transformed in 4D volumes (skeletons) in which the fourth dimension represents the subject ID. The final step of the analysis consisted in feeding the 4D volumes in to a voxel-wise cross-subject general linear model, with PRS and diagnostic group as independent variables and FA and MD as dependent variables. Main effects and interaction effects were inferred. No significant correlation between PRS and FA or MD was found, either as a main effect or a diagnosis-dependent effect. Nevertheless, small main effect trends revealed a positive association of PRS with FA in uncinate fasciculus, corpus callosum, middle cerebellar peduncle and anterior thalamic radiation. In addition, other main effect trends of PRS revealed a negative association with FA in the superior longitudinal fasciculus, posterior and anterior thalamic radiation, corticopontine tract, middle cerebellar peduncle and cingulum. Regarding MD, positive main effect trends of PRS on this measure were found in the inferior longitudinal fasciculus and posterior corona radiata, while negative effects were found in inferior cerebellar peduncle. There were also no significant diagnosis by PRS interactions on FA, albeit, trend level effects showed that PRS and FA have higher correlation in REL than SZ patients in the middle cerebellar peduncle. Main effects of diagnosis on FA and MD are reported for completeness. In sum, I have found no evidence to support a role for the current SZP RS genetic risk factors in contributing for the known decreased FA and increased MD in SZ and BD. However, given the small penetrance of SZ genetic variants, and the small sample size of this group, this approach should be replicated in a larger sample size, before an effect of PRS on WM microstructure can be excluded.